Range Based Flushing Mechanism

1. An apparatus to facilitate memory flushing, comprising: a cache memory; one or more processing resources; tracker hardware to dispatch workloads for execution at the processing resources and to monitor the workloads to track completion of the execution; range based flush (RBF) hardware to process RBF commands and generate a flush indication to flush data from the cache memory, wherein the flush indication comprises flush command attributes; and a flush controller to receive the flush indication and perform a flush operation to discard the data from the cache memory at an address range provided in the flush indication, wherein the flush controller prioritizes flush operations based on the flush command attributes.

2. The apparatus of claim 1 , wherein the tracker hardware dispatches a first set of workloads, wherein each workload in the first set of workloads comprises a first group identifier.

3. The apparatus of claim 2 , wherein the tracker hardware tracks the completion of each of the workloads.

4. The apparatus of claim 3 , wherein the tracker hardware transmits an indication to the RBF hardware including the first group identifier indicating that the first set of workloads has completed the execution.

5. The apparatus of claim 4 , wherein the RBF hardware processes a first RBF command comprising the first group identifier upon receiving the indication.

6. The apparatus of claim 5 , wherein the RBF hardware transmits a first flush indication to the flush controller to perform a first flush operation.

7. The apparatus of claim 6 , wherein the RBF hardware tracks the completion of the first flush operation.

8. The apparatus of claim 7 , wherein the RBF hardware tracking the completion of the first flush operation comprises receiving an indication from the flush controller.

9. The apparatus of claim 1 , wherein the flush controller further monitors bandwidth utilization of the cache memory and automatically performs the flush operations upon determining that the bandwidth utilization is below a predetermined threshold.

10. A method to facilitate memory flushing, comprising: dispatching a first set of workloads for execution; tracking the execution of the first set of workloads to determine whether the first set of workloads has completed the execution, wherein the first set of workloads is tracked via a first group identifier associated with each workload in the first set of workloads; and processing a first range based flush (RBF) command to generate a flush indication to flush data from a cache memory upon a determination that the first set of workloads has completed the execution, wherein the flush indication comprises flush command attributes; and performing a flush operation to discard the data from the cache memory at an address range provided in the flush indication, wherein the flush controller prioritizes flush operations based on the flush command attributes.

11. The method of claim 10 , wherein the first RBF command comprises the first group identifier.

12. The method of claim 11 , further comprising tracking the completion of the first flush operation.

13. The method of claim 12 , wherein tracking the completion of the first flush operation comprises receiving an indication of a status of the first flush operation.

14. A graphics processing unit (GPU), comprising: a memory device; a cache memory; one or more processing resources; tracker hardware to dispatch workloads for execution at the processing resources and to monitor the workloads to track completion of the execution; range based flush (RBF) hardware to process RBF commands and generate a flush indication to flush data from the cache memory, wherein the flush indication comprises flush command attributes; and a flush controller to receive the flush indication and perform a flush operation to discard the data from the cache memory at an address range provided in the flush indication, wherein the flush controller prioritizes flush operations based on the flush command attributes.

15. The GPU of claim 14 , wherein the tracker hardware dispatches a first set of workloads, wherein each workload in the first set of workloads comprises a first group identifier.

16. The GPU of claim 15 , wherein the tracker hardware tracks the completion of each of the workloads.

17. The GPU of claim 16 , wherein the tracker hardware transmits an indication to the RBF hardware including the first group identifier indicating that the first set of workloads has completed the execution.